VQC: Computer Vision Quality Control

Developed as an R&D initiative for Balin Tech, the VQC project addresses the critical end-of-line inspection for aluminum cylinder heads. The goal was to eliminate human error in detecting missing machined holes and surface defects such as scratches, cavities, or porosity.

The system creates a dynamic comparison model: it takes an 'Ideal Reference' input and maps subsequent products using reference holes for precise alignment. This allows the software to support product diversity (tested on 3 types) and identify non-conformities with high precision.

• Reflective Surface Control: Overcoming specular reflections on aluminum surfaces through engineered uniform lighting design.
• Dynamic Reference Mapping: Developing an algorithm to align and map production parts to a "Master Sample" using geometric reference points.
• Multi-Product Support: Creating a flexible software architecture capable of handling 3 different cylinder head models without code changes.
• Defect Classification: Distinguishing between critical casting defects (cavities) and harmless surface anomalies.